Hacker News Reader

Comment by timmg | Hacker News Reader

Quite honestly, I love what MS has done with JavsScript with Typescript. It would be great if they could do the same for C.

I guess the hard part will be the fact that all your dependent libraries will (likely) still be unsafe.

View on HN

I'm a bit confused, at the start they talk about C++, but some of the code examples look like C. Do those checks also work in the C compiler? (the difference is kinda important, because unlike Clang, the MSVC C++ compiler doesn't have enough non-standard language extensions to compile C code that was written in the last 20 years).

They talk about C++, because that is what mostly matters to Visual C++ customers, however /analyse mode also works in C code.

https://learn.microsoft.com/en-us/cpp/code-quality/code-anal...

I'm a bit confused. Didn't Herb Sutter essentially fork clang and extended the codebase in the beginning? Is microsoft working with that fork and (most importantly) re-contributing?

If you mean cpp2, that has nothing to do with Microsoft, it's Herb's personal project.

No, his lifetime work, e.g. https://herbsutter.com/2018/09/20/lifetime-profile-v1-0-post...

No that was a proof of concept for a paper that went nowhere.

VC++ analysis builds up on SAL infrastructure.

https://learn.microsoft.com/en-us/cpp/code-quality/understan...

"Closing the Gap between Rust and C++ Using Principles of Static Analysis - CppCon 2020"

https://www.youtube.com/watch?v=_pQGRr4P16w

There is also a clang work on lifetimes, but little progress has happened since 2019.

"Lifetime analysis for everyone - CppCon 2019"

https://www.youtube.com/watch?v=d67kfSnhbpA

SAL, as I understand it, is Visual Studio/Microsoft only? Do you have any idea if this work will make it to clang/gcc/g++?

SAL is deliberately designed such that it can be defined as a bunch of macros that expand to whatever compiler-specific magic the implementation needs (e.g. MSYS64 doesn't support it at all, but they define all those macros as no-op, so code still compiles). So in principle it could work across implementations.

I'm begging them to do the same for Python.

They made pyright which is pretty rad

Yeah it's pretty great, but still hamstrung by the inherent limitations of the language itself. Would love to see a new language on top of Python that treats types as first-class constructs.

Why "on top of Python"? The runtime is slow as hell, limited and actually very aged.

But there are better modern alternatives.

https://docs.scala-lang.org/scala3/book/scala-for-python-dev...

Types in Scala are not only "first-class". You can even program with them like in TS.

https://blog.rockthejvm.com/type-level-programming-scala-3/

(Type level functions are correctness proves, BTW.)

Enjoy!

I guess I should have mentioned that Scala 3, with its new syntax, look almost like "Python with types". That's why I've mentioned it.

Reply on news.ycombinator.com